luigidemasi opened a new issue, #34: URL: https://github.com/apache/camel-kit/issues/34
> Imported from [luigidemasi/camel-kit#61](https://github.com/luigidemasi/camel-kit/issues/61), opened by [luigidemasi](https://github.com/luigidemasi) on 2026-05-12. > Closed by [luigidemasi/camel-kit#70](https://github.com/luigidemasi/camel-kit/pull/70), merged on 2026-05-13. --- ## Description Restructure the camel-kit pipeline to support file-based handoff between stages and clean-context invocability. Each pipeline step reads its input from files and writes its output to files, enabling session resilience, multi-session development, and CI/CD integration. **Related:** luigidemasi/camel-kit#60 (subagent composition model) **ADR:** ADR-0031 (pipeline-file-based-handoff) **Research sources:** [spec-kit](https://github.com/github/spec-kit), spec-kit-refine extension ## Pipeline Architecture ``` brainstorm/migrate → plan → execute [+verify subagent] → validate ``` ## Directory Structure ``` docs/camel-kit/ 001-order-processing/ design-spec.md implementation-plan.md execution-report.md validation-report.md .camel-kit/pipeline.json → {"activePipeline": "001-order-processing"} ``` ## Settled Decisions | Decision | Resolution | |----------|-----------| | **Motivation** | Session resilience + CI/CD (not context management — orthogonal to ADR-0029/0030) | | **Dual-mode invocation** | Chained mode (auto-invoke, HARD-RULEs preserved) + standalone mode (reads from files) | | **Verify** | Moves to Internal tier, dispatched as subagent by execute. Future `/camel-debug` for ad-hoc debugging | | **Validate** | Promoted to Tier 1 as final pipeline step. Critical findings = PAUSE regardless of oversight level | | **Artifact location** | `docs/camel-kit/<PIPELINE_ID>/` with generic filenames — version-controlled, PR-visible | | **PIPELINE_ID** | `NNN-slug` (e.g., `001-order-processing`). Sequential or timestamp, configurable | | **Stage detection** | Artifact presence — which files exist. No `currentStage` tracking (spec-kit pattern) | | **Pipeline state** | Minimal: `.camel-kit/pipeline.json` → `{"activePipeline": "001-order-processing"}` | | **Re-iteration** | Allowed via `/camel-brainstorm <PIPELINE_ID>`. Staleness markers (`⚠️ STALE`) embedded in downstream docs (spec-kit-refine pattern). Manual cascade; camel-ship auto-detects on resume | | **Decision rationale** | Structured Rationale + Constraint fields enforced per component/EIP decision in design specs | | **Verify iteration log** | Structured log (`.camel-kit/verify-log.md`) as audit trail | ## Updated Tier Classification | Tier | Skills | Role | |------|--------|------| | Meta | `camel-start` | Routes to the right pipeline | | Tier 1 (Pipeline) | `brainstorm`, `migrate`, `plan`, `execute`, `validate` | The 5 pipeline steps | | Tier 2 (Power) | `ship`, `knowledge` | Standalone utilities | | Internal | `design`, `implement`, `test`, `verify` | Guide libraries / subagent-only | ## Oversight Matrix **Execute (includes verify):** | Outcome | `always` | `smart` | `never` | |---|---|---|---| | Tasks done, verification PASS | PAUSE (present report) | PAUSE (present report) | AUTO-PROCEED | | Tasks done, verification FAIL | PAUSE | PAUSE | AUTO-FIX | **Validate:** | Outcome | `always` | `smart` | `never` | |---|---|---|---| | No Critical findings | PAUSE (present report) | DONE | DONE | | Critical findings | PAUSE | PAUSE | PAUSE | ## Implementation Steps - [ ] Create pipeline ID generation script (scan `docs/camel-kit/`, find max, increment, create directory) - [ ] Implement `docs/camel-kit/<PIPELINE_ID>/` directory structure with generic filenames - [ ] Add `.camel-kit/pipeline.json` with minimal `activePipeline` field - [ ] Add standalone mode detection to brainstorm, plan, execute, validate skills (read from files when no conversation context) - [ ] Move verify to Internal tier (`user_invocable: false`, remove slash command) - [ ] Update execute to dispatch verify as subagent and include verification in execution report - [ ] Promote validate to Tier 1 — add to pipeline, update oversight matrix - [ ] Merge verify rows into execute in oversight matrix - [ ] Add Critical-findings-always-PAUSE rule to validate oversight - [ ] Implement staleness markers (`⚠️ STALE`) in brainstorm amend flow - [ ] Update camel-ship `--resume` to detect staleness markers and re-run stale stages - [ ] Enforce structured Rationale + Constraint fields in design-assembly guide - [ ] Add structured iteration log to verify skill - [ ] Update camel-start decision tree with new tier classification - [ ] Update `--start-from` stage mapping (remove verify stage, add validate) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
